Eval only and batch normalization

jean_zay/submitit/understanding-unets/eval_only_unet_training.py

@@ -0,0 +1,44 @@
+from learning_wavelets.evaluation_scripts.exact_recon_unet_evaluate import evaluate_unet
+
+from jean_zay.submitit.general_submissions import eval_grid
+
+
+job_name = 'exact_recon_unet_training'
+n_epochs = 100
+base_n_filters = 4
+n_layers = 4 
+non_linearity = 'relu'
+bn = True
+exact_recon = True
+run_id = 'ExactReconUnet_4_bsd500_0_55_None_1620730822'
+n_gpus = 1
+possible_std_dev = [0.0001, 5, 15, 20, 25, 30, 50, 55, 60, 75]
+
+base_parameters = dict(
+    n_epochs=n_epochs,
+    base_n_filters=base_n_filters,
+    n_layers=n_layers,
+    non_linearity=non_linearity,
+    bn=bn,
+    exact_recon=exact_recon,
+)
+
+parameters = [
+    base_parameters,
+]
+
+res_all = eval_grid(
+    job_name,
+    evaluate_unet,
+    parameters,
+    run_ids=[run_id],
+    to_grid=False,
+    timeout=1,
+    n_gpus=n_gpus,
+    project='learnlets',
+    params_to_ignore=['batch_size'],
+    noise_std_test=possible_std_dev, 
+)
+
+print('Results')
+print(res_all)

jean_zay/submitit/understanding-unets/exact_recon_unet_training.py

@@ -10,6 +10,8 @@
 base_n_filters = 4
 n_layers = 4 
 non_linearity = 'relu'
+bn = True
+exact_recon = True
 n_gpus = 2
 possible_std_dev = [0.0001, 5, 15, 20, 25, 30, 50, 55, 60, 75]
 
@@ -19,6 +21,8 @@
     base_n_filters=base_n_filters,
     n_layers=n_layers,
     non_linearity=non_linearity,
+    bn=bn,
+    exact_recon=exact_recon,
 )
 
 parameters = [